Comparing Poisson Rates When Data Are Subject to Underreporting

Key Points

• Underreporting compromises the statistical power of the conditional test when comparing Poisson rates under a standard binomial assumption.
• Theoretical model development yields a large sample normal test alongside a parametric bootstrap test, correcting for missing counts using a double sample.
• These corrected testing procedures maintain inferential validity across count data, though application remains constrained by the accuracy of the double sample.